CN107680076A - Stereo-picture comfort level contrast range measuring method based on salient region - Google Patents
Stereo-picture comfort level contrast range measuring method based on salient region Download PDFInfo
- Publication number
- CN107680076A CN107680076A CN201710729853.2A CN201710729853A CN107680076A CN 107680076 A CN107680076 A CN 107680076A CN 201710729853 A CN201710729853 A CN 201710729853A CN 107680076 A CN107680076 A CN 107680076A
- Authority
- CN
- China
- Prior art keywords
- stereo
- contrast
- picture
- image
- dimensional
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
- G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to image processing field, it is contemplated that providing the quantitative criterion of more reasonable for the making of stereo content.The technical solution adopted by the present invention is the stereo-picture comfort level contrast range measuring method based on salient region, and step is as follows:1) extraction of three-dimensional saliency map:Two dimensional image notable figure is combined to obtain three-dimensional saliency map with stereo image parallax figure;2) the significantly acquisition of stereo-picture:Processing is optimized to three-dimensional saliency map, three-dimensional saliency map optimizes to obtain notable stereo-picture through fuzzy theory and mask;3) subjective experiment:Notable stereo-picture is subjected to contrast variation for step-length with 0.1 and carries out subjective experiment;4) comfortable contrast matching figure and comfortable contrast disparity map are obtained.Present invention is mainly applied to image procossing.
Description
Technical field
The invention belongs to image processing field, is related to the Elements research for influenceing stereo-picture viewing comfort level, especially relates to
A kind of and comfortable contrast Quantitative research method based on stereo-picture marking area.
Background technology
Stereo display technique can also make beholder produce different journeys while visual experience on the spot in person is brought to user
The discomfort of degree.Visual fatigue, visual sensitivity reduce, and the uncomfortable phenomenons of vision such as eyes regulating power declines are considered as
It is major obstacle of the 3D products in market segment Quick Extended[1][2].Research shows to cause the uncomfortable principal element bag of vision
Include:Four aspects such as viewing environment, stereoscopic display device, the Physiological Psychology situation of people and stereo content, wherein in comfortable solid
The shortage of appearance is to restrict one of an important factor for stereoscopic imaging technology is popularized[3].The factor for influenceing stereo content comfort level has perhaps
It is more, such as contrast, brightness, colourity, saturation degree, crosstalk and parallax etc.[4].It is significantly special that the present invention combines human eye vision first
Property, influence of the quantitative study contrast factor to visual comfort, obtaining contrast factor influences the quantitative criterion of comfort level.
Domestic and international experts and scholars have carried out widely studied to the factor for influenceing stereo-picture comfort level.Document [5] [6] [7]
The relation between visual comfort and stereo image parallax size, parallax distribution, stereo content image pickup method is have studied respectively.
Influence of the colourity of document [8] quantitative study stereo-picture to stereo-picture comfort level, shows the colour difference when left and right view
Different stereo-picture comfort level when exceeding certain thresholding reduces.Document [9] explicitly points out, and it is larger to watch binocular disparlty for a long time
Stereo content can cause visual fatigue and induce eye disease.Document [10] shows that the Luminance Distribution of stereo-picture can influence to stand
The comfort level held in vivo.The factors such as document [11] comprehensive study parallax, the complexity of image background, contrast are to stereogram
As the qualitative effect of quality.
The factor that stereo content comfort level is influenceed due to quantitative study is taken time and effort, comfortable to stereo-picture both at home and abroad at present
The research of degree is mostly qualitatively, and still without forming unified clear and definite standard.In order to more accurately judge relaxing for stereo-picture
Appropriateness, the making for stereo content provide direct basis, it is necessary to carry out quantitative study.Document [12] quantitative study contrast
Influence of the factor to stereo-picture comfort level, research method of this literary grace based on view picture stereo-picture, does not take into full account people
The visual characteristic of eye.Numerous researchs show, when watching stereo content, the notice of beholder can be at once by aobvious in scene
Write region to attract, it is aobvious that subsequent human visual system (Human Visual System, HVS) controls the crystalline lens of human eye to focus on
Write object so that human eye often only observes signal portion therein rather than entire image, this phenomenon when observing image
Referred to as vision attention, vision noticing mechanism comfortably have a major impact to human eye vision[13], therefore the present invention combines vision first
Attention mechanism, pass through influence of a large amount of subjective experiment quantitative study contrast factors to stereo-picture comfort level.The work is first
The method being combined using plane notable figure and stereoscopic parallax figure obtains three-dimensional saliency map, by the three-dimensional significance after optimization
Figure is combined to obtain notable stereo-picture with source images, and utilizes the three-dimensional significantly algorithm of eye tracker experimental verification to extract notable area
The correctness in domain;Then contrast linear change is carried out to left and right view and obtains distorted image progress subjective experiment.To testing number
It is fitted according to least square method, obtains the comfortable contrast matching figure of notable stereo-picture and comfortable contrast disparity map,
And each several part in figure is analyzed.Contrast verification and the result of test checking show that what the method according to the present invention obtained relaxes
Suitable contrast range more meets human eye vision subjective observation result, and more scientific reasonable and letter is provided for the making of stereo content
Single easy quantitative criterion.
The content of the invention
For overcome the deficiencies in the prior art, it is contemplated that providing the quantitative mark of more reasonable for the making of stereo content
It is accurate.The technical solution adopted by the present invention is the stereo-picture comfort level contrast range measuring method based on salient region,
Step is as follows:
1) extraction of three-dimensional saliency map:Two dimensional image notable figure is combined to obtain three-dimensional significance with stereo image parallax figure
Figure;
2) the significantly acquisition of stereo-picture:Processing is optimized to three-dimensional saliency map, three-dimensional saliency map is through fuzzy original
Reason and mask optimize to obtain notable stereo-picture;
3) subjective experiment:Notable stereo-picture is subjected to contrast variation for step-length with 0.1 and carries out subjective experiment;
4) comfortable contrast matching figure and comfortable contrast disparity map are obtained.
The extraction of three-dimensional saliency map comprises the concrete steps that:GBVS (Graph-based visual are used first
Saliency) the plane notable figure SM of right view is calculated in algorithmR(x, y), then obtained using Fast Stereo Matching Algorithm
Disparity map d on the basis of right viewR(x, y), finally visual saliency map and right anaglyph are added using linear mode
Power obtains three-dimensional saliency map I (x, y), such as formula (1):
I (x, y)=w1dR(x,y)+w2SMR(x,y) (1)
In formula:I (x, y) is three-dimensional saliency map;w1And w2To weight proportion, and w1+w2=1.
The acquisition detailed process of notable stereo-picture is that domain X is three-dimensional saliency map, by domain pixel (x,
Y) it is divided into two classes, marking area A and non-significant region B, A and B are one of X and divided, i.e. A ∪ B=X,
The pixel is under the jurisdiction of the degree of marking area, pixel in the gray value expression original image of three-dimensional saliency map
The gray value of (x, y) is L, and this is to marking area A degree of membership
For obtain marking area A need carry out deblurring operation, using the method for Threshold segmentation obtain mask images M (x,
Y), as shown in formula (2), segmentation threshold T is determined by maximum between-cluster variance method:
For any pixel point (x, y) in three-dimensional saliency map I (x, y), if (x, y) is to the degree of membership of marking area
More than threshold value T, then the pixel belongs to vision significance region, the white portion in corresponding mask images M (x, y), otherwise belongs to
In black region, mask image M (x, y) is optimized using morphologic opening and closing operation to obtain optimization mask image M'(x,
y);
Finally optimization mask images are multiplied with original visual point image to obtain notable stereo-picture.
Contrast variation carries out comprising the concrete steps that for subjective experiment, linear using contrast after obtaining notable stereo-picture
Converter technique, different grades of contrast distortion is carried out to left and right view, obtains a series of stereo-picture to be measured of distortions, to figure
During as carrying out contrast variation, the gray value of note image original image vegetarian refreshments is In, is calculated by contrast variation's formula (3)
The gray value Out of pixel after conversion:
Out=Avg+ (In-Avg) × (1- α) (3)
Wherein Avg represents the average gray value of whole pictures, and α ∈ [- 1,1] are the setting contrast factor, control output picture
Vegetarian refreshments and the relation of input pixel gray value, the conversion step-length for taking α is 0.1, obtains 21 contrast variation's straight lines, corresponding
The contrast distortion of 21 kinds of grades, transverse axis are the gray value In of input pixel, and the longitudinal axis is to be obtained after pixel spends conversion by contrast
The gray value Out arrived, every contrast variation's straight line all passing points (Avg, Avg) and on (Avg, Avg) symmetrically, ensure that
The mean flow rate of picture is kept constant before and after contrast variation.
Obtain comfortable contrast matching figure and comfortable contrast disparity map comprises the concrete steps that, using least square piecewise linearity
Fitting process is handled 4 groups of experimental datas, is obtained 4 groups of stereo-pictures and is comfortably normalized contrast matching figure MM (Matching
Map contrast disparity map DM (Difference Map)) and is comfortably normalized.
The features of the present invention and beneficial effect are:
View-based access control model attention mechanism of the present invention, contrast factor is to stereo-picture by a large amount of subjective experiment quantitative studies
The influence of comfort level, compared to the experiment for not considering vision noticing mechanism, the present invention has preferably reacted the visual impression of human eye
By, and the comfortable contrast matching figure of stereo-picture and the comfortable contrast disparity map of stereo-picture have been obtained, such as Fig. 7 and Fig. 8 institutes
Show.The two result figures embody influence of the contrast change of binocular stereo image to human eye comfortableness, can be according to this hair
Bright experimental result judges whether stereo-picture is comfortable, compensate for the shortcomings that subjective experiment is wasted time and energy well;To result figure
Comprehensive labor, the also making for stereo content and processing provide reference and foundation.Compared to not considering in document [12]
The experiment of vision noticing mechanism, present invention gained comfort standard more meet human-eye visual characteristic, can be that the making of stereo content carries
For more rationally more feasible quantitative criterion.
Brief description of the drawings:
Fig. 1 width source stereo-picture.
Fig. 2 algorithm flow charts.
The notable stereogram extraction process figures of Fig. 3.
Fig. 4 eye trackers test notable thermal map.
Fig. 5 contrast linear change schematic diagrames.
The comfortable contrast matching figure of Fig. 6 stereo-pictures.
The comfortable contrast disparity map of Fig. 7 stereo-pictures.
Fig. 8 experimental results of the present invention and document [12] Comparative result.
Fig. 9 sources stereo-picture.
Embodiment
For the making of stereo content, to provide quantitative criterion be that stereo product popularizes key issue urgently to be resolved hurrily.It is of the invention first
It is secondary with reference to vision noticing mechanism, by a large amount of subjective experiment quantitative study contrast factors to stereo image vision comfort level
Influence.The present invention first combines plane notable figure and stereoscopic parallax figure obtains three-dimensional saliency map, recycles fuzzy membership and covers
Film optimizes to it obtains final notable stereo-picture, and verifies its reasonability with eye tracker;Then left and right view is contrasted
Degree is converted and carries out subjective experiment, and the comfortable contrast matching figure and disparity map of notable stereo-picture are obtained after data screening.It is real
Result is tested to show:The poor contrast exclusive-OR gate limit value of left and right view changes as the contrast value of left view is different, and left and right regards
Figure contrast difference can not be excessive, and maximum, minimum difference value is respectively 1.97 and -2.40.The comfortable contrast range of experiment gained
The comfort level of stereo-picture is reflected well, and the accuracy of confirmatory experiment provides more up to 95% for the making of stereo content
The quantitative criterion of reasonable.
Experiment flow of the present invention is the extraction of 1) three-dimensional saliency map.Two dimensional image notable figure and stereo image parallax figure knot
Conjunction obtains three-dimensional saliency map.2) the significantly acquisition of stereo-picture.Processing is optimized to three-dimensional saliency map, it is three-dimensional notable
Degree figure optimizes to obtain notable stereo-picture through fuzzy theory and mask.3) subjective experiment.By notable stereo-picture with 0.1 for step
It is long to carry out contrast variation and carry out subjective experiment.4) comfortable contrast matching figure and comfortable contrast disparity map are obtained.Made
Source stereo-picture is as shown in figure 1, experiment flow figure is as shown in Figure 2.
Detailed analysis will be carried out to each step below:
1. the acquisition of three-dimensional saliency map
According to human visual system's characteristic, human eye is in non-uniform Distribution to the notice of scene different piece, most of note
Power of anticipating concentrates marking area in the scene, less to the non-significant region concern in scene[14].Therefore, can use significantly vertical
The comfort level of body image reflects the comfort level situation of view picture stereo-picture, to improve test accuracy rate and to reduce calculating complicated
Degree.The parallax characteristic and image space frequency of stereo-picture can all influence visual comfort[15], therefore shown by two dimensional image
Write figure and stereo image parallax figure obtains three-dimensional saliency map.
Majority's dominant eye is right eye, therefore the present invention uses GBVS (Graph-based visual saliency) first
Algorithm[16]The plane notable figure SM of right view is calculatedR(x,y).Then Fast Stereo Matching Algorithm is used[17]Obtain with the right side
Disparity map d on the basis of viewR(x,y).Finally visual saliency map and right anaglyph are weighted using linear mode
To three-dimensional saliency map I (x, y), such as formula (1):
I (x, y)=w1dR(x,y)+w2SMR(x,y) (1)
In formula:I (x, y) is three-dimensional saliency map;w1And w2To weight proportion, and w1+w2=1;W is taken in the present invention1=w2
=0.5.
2. the acquisition of notable stereo-picture
In obtained three-dimensional saliency map, the edge blurry of salient region of image can not be direct by source images according to this figure
Marking area and non-significant region are divided into, therefore is subordinate to topology degree using fuzzy mathematics to describe each pixel in image
Significance degree[18], three-dimensional saliency map is optimized, obtains a width binary mask image M[19].The pixel value of pixel is 1
Show that the point in former stereo-picture belongs to marking area, otherwise belong to non-significant region.Detailed process is as follows:
Domain X is three-dimensional saliency map, and the pixel (x, y) in domain is divided into two classes, marking area A and non-significant area
Domain B, A and B are one of X divisions, i.e. A ∪ B=X, For empty set.
The pixel is under the jurisdiction of the degree of marking area, pixel in the gray value expression original image of three-dimensional saliency map
The gray value of (x, y) is L, and this is to marking area A degree of membership
Need to carry out deblurring operation to obtain marking area A[18].The present invention obtains mask using the method for Threshold segmentation
Image M (x, y)[19], as shown in formula (2).Segmentation threshold T is determined by maximum between-cluster variance method[20]。
For any pixel point (x, y) in three-dimensional saliency map I (x, y), if (x, y) is to the degree of membership of marking area
More than threshold value T, then the pixel belongs to vision significance region, the white portion in corresponding mask images M (x, y), otherwise belongs to
In black region.In order to remove the burr at mask image edge and cavitation, using morphologic opening and closing operation to mask figure
As M (x, y) optimizes to obtain optimization mask image M'(x, y).
Finally optimization mask images are multiplied with original visual point image to obtain notable stereo-picture, as shown in Fig. 3 (a-g).
Fig. 3 sets forth the right viewpoint of 4 groups of stereo-pictures, disparity map, plane notable figure, three-dimensional saliency map, mask image, optimization
Mask image and notable stereo-picture.
In order to verify whether the salient region of significantly stereo-picture obtained by inventive algorithm is accurate, the present invention is moved using eye
Instrument carries out subjective measurement to the marking area of source stereo-picture.Calibration experiments and reference measurement are first carried out before experiment every time, works as quilt
Examination person tests when being less than 1 degree to the observation error for calibrating datum mark.Presented at random per pictures and the presentation time is 5s.
The thermal map for the interesting image regions measured according to eye tracker, found with the notable stereo-picture regional correlation of the present invention, two
The marking area that person is proposed is basically identical, and Fig. 4 is the thermal map for the interesting image regions that eye tracker is measured.
3. contrast variation carries out subjective experiment
After obtaining notable stereo-picture, using contrast linear transformation method[12], it is different grades of right that left and right view is carried out
Than degree distortion, a series of stereo-picture to be measured of distortions is obtained.When carrying out contrast variation to image, image original pixels are remembered
The gray value of point is In, and the gray value Out of pixel after conversion is calculated by contrast variation's formula (3):
Out=Avg+ (In-Avg) × (1- α) (3)
Wherein Avg represents the average gray value of whole pictures, and α ∈ [- 1,1] are the setting contrast factor, control output picture
Vegetarian refreshments and the relation of input pixel gray value.It is 0.1 that present invention experiment, which takes α conversion step-length, obtains 21 contrast variations
Straight line, the contrast distortion of corresponding 21 kinds of grades, as shown in Figure 5.Transverse axis is the gray value In of input pixel, and the longitudinal axis is picture
Vegetarian refreshments spends after conversion obtained gray value Out by contrast.Every contrast variation's straight line all passing points (Avg, Avg) and on
(Avg, Avg) symmetrically, ensure that the mean flow rate of picture before and after contrast variation keeps constant.
4. obtain experimental result picture
Using least square Sectioned linear fitting method[21]4 groups of experimental datas are handled, 4 groups of stereo-pictures is obtained and relaxes
Suitable normalization contrast matching figure (Matching Map, MM) and comfortable normalization contrast disparity map (Difference
Map, DM), as shown in Figure 6,7.
Experimental facilities and material
The three-dimensional display used in experiment is 22 inches of polarization type 3D three-dimensional displays of AOC (AOC) company, uses moral
The correctness of the iView X RED eye trackers checking marking area algorithm extraction of SM companies of state.
Experimental material selects the stereoscopic image data storehouse that advanced science and technology institutes of South Korea provides as stereo-picture sample
Data acquisition system[22].The stereo-picture storehouse is made up of the stereo pairs that 120 pairs of resolution ratio are 1920 × 1080, covers personage,
Polytype image such as plant, building.Four width source stereo-picture ball.bmp in this experimental selection stereo-picture storehouse,
Plant.bmp, orange.bmp and baby.bmp, as shown in figure 1, this 4 width source stereo-picture third dimension is strong, visual comfort
Height, contrast value are distributed in different gradients, have preferably representative.
Experimental situation and subject
According to ITU subjective assessment standards[23], when being tested the illumination brightness of surrounding environment have to be lower than 3lux, for
22 inches of full HD displays, viewing distance should be three times of screen height.
To ensure the accuracy of conclusion, the work(that 30 selected subjects have passed through regular ophthalmologic hospital is tested
It can check.17 males, 13 women are selected, wherein have 10 people of priori, 20 people of no priori, all subjects
Dominant eye be all right eye.
Subjective assessment
Subjective assessment is the important evidence of paper experiment quantitative study.Before experiment, first experimental science is introduced to subject
Subjective assessment process be quantitative study of the present invention basis.Need to introduce experimental evaluation method to subject before evaluation and carry out
Pre-training, pre-training image should have completeness, enable the affiliated grade of subject's correct understanding image, and length of testing speech must not surpass
Spend 0.5 hour.Double stimulations damage Scale Method (Double-Stimulus Impairment that subjective experiment is suggested using ITU
Scale, DSIS) method, grading system and corresponding damage term are as shown in table 1[24]。
After the completion of subjective experiment, in order to exclude the knowledge background because of 30 subjects, the experiment factor such as motivation influence, it is necessary to
Exceptional value in knock-out experiment data[23].The present invention proposes two-stage data screening method, uses be based on average and confidence first
The Grubbs' test method in section[24], the exceptional value in all subjective scores obtained for rejecting certain width image;Then adopt
With the method recommended in standard (ITU-R) BT.500, for detecting whether the very big subject of effort analysis be present, if in the presence of if
Superseded this, which is tested and rejects it, provides all fractions.Two-stage data screening strict guarantee experiment the data obtained of the invention
Objectivity.
Statistical average is carried out to the experimental data after screening and obtains the mean subjective score value (Mean of a width stereo-picture
Opinion Score, MOS), the MOS values of image are higher to show that comfort level is better, it is specified that the stereo-picture of MOS >=4 relaxes for vision
Suitable stereo-picture[25]。
With reference to technical scheme, experimental result that labor this method obtains:
First, to the analysis of Fig. 6 Fig. 7 experimental results
The euphorosia region division that piecewise fitting straight line is surrounded is 4 parts, the analysis such as table 2 to Fig. 6 each several parts:
(1) from Fig. 6 and table 2, R1 regions are asymmetric on straight line y=x and above straight line, show R1 regions
Right view contrast is higher than left view.Because right eye is dominant eye, right view acts on larger, left view contrast mistake during binocular fusion
When low, from Fechner antinomys, the low right of left view point can be suppressed to a certain extent by suitably increasing the contrast value of right viewpoint
Than influence of the degree to visual comfort, now major-minor eye works.R2 regions are symmetrical on straight line y=x, illustrate acceptor
The influence of auxiliary eye is smaller.R3, R4 region are asymmetric on y=x, below straight line, show that left view contrast regards higher than the right side
Figure, because when left view contrast is too high, the appropriate contrast value for reducing right view can suppress left view to a certain extent
The high-contrast of figure stimulates the influence to visual comfort, and major-minor eye works.
(2) in Fig. 6, R1 regions left view contrast value is very low, and the image texture that left eye perceives is unintelligible, details mould
Paste, when combining to form stereo-picture with right view, human eye can be it is observed that view clean mark another view almost have
Textured, binocular disparlty is excessive to be difficult to binocular fusion, therefore as left view contrast value gradually increases, binocular disparlty
Reduce, zone of comfort gradually increases.The value of R2 regions or so view normalization contrast is stood all near 1.5 equivalent to source
Body image carries out contrast enhancing by a small margin, and left and right view clean mark, visual quality is preferable, and contrast difference is little,
Easily carry out binocular fusion and obtain the high stereo-picture of comfort level, therefore R2 region areas are larger.R3 regions left view contrast
Higher, image engenders cross-color, the relatively low region loss in detail of brightness, the fuzzy phenomenon of texture, therefore R3 comfort zones
Domain is gradually reduced.R4 region contrasts are very high, and it is uncomfortable that high-contrast stimulation causes vision.When right viewpoint contrast compared with
Small, two viewpoint differences are excessive, and eyes can not comfortably melt picture;When the contrast of right viewpoint is excessive, the image that binocular fusion obtains
Contrast is too high, and visual comfort is still bad, therefore in R4 regions, zone of comfort, which is gradually reduced and shunk, to be tended to close.
(3) as shown in Figure 7, when the contrast difference of left view point, the right left view point contrast difference of comfortable stereo-picture
Threshold value it is also different.A given width left view and its contrast, it is comfortably three-dimensional a width to be synthesized according to Fig. 7 determinations
Image, more significantly contrast enhancings can be carried out on the basis of left view or are weakened.First and last, right left view contrast
The minimum value for spending difference is -2.40, and the maximum of right left view contrast difference is 1.97, two view contrasts of stereo-picture
The difference of degree can not be excessive, otherwise can reduce the viewing comfort level of stereo-picture.Due to Fig. 7 symmetrical, figures that are not in relation to leading diagonal
7 are also not in relation to horizontal axial symmetry, illustrate that major-minor eye has not when investigation and comparison degree factor is to binocular stereo image comfort level
The influence that can ignore, if the dominant eye of subject is left eye, the zone of comfort for testing to obtain is symmetrical on y=x with Fig. 6.Comfortably
In region, region area of the contrast more than 1 is less than 1 region area much larger than contrast, illustrates contrast increase to solid
The influence of image comfort level reduces the influence to comfort level much smaller than contrast.
(4) it is the analysis to euphorosia region more than, the uncomfortable reason of stereo-picture is caused below in conjunction with Fig. 6 analyses
And solves method.The upper left corner and the lower right corner in figure are non-zone of comfort, and it is excessive to be primarily due to left and right viewpoint contrast difference,
Eyes can not comfortably melt picture, watch the excessive stereo-picture of binocular disparlty for a long time or three-dimensional video-frequency easily causes visual fatigue, lure
Eye disease is sent out, the contrast of right viewpoint now should suitably be decreased or increased, be easy to human eye binocular fusion to obtain comfort chart
Picture;The lower left corner and the upper right corner in figure are non-zone of comfort, although the contrast for being primarily due to left and right viewpoint is close, are easy to double
Mesh merges, but because the contrast of two viewpoints is all too low or too high, the stereo image quality for merging to obtain is bad, causes
Vision is uncomfortable, when making binocular stereo image it is noted that using dominant eye image adjustment effect, such as dominant eye view can be used
The high-contrast of smaller another view of contrast inhibition stimulates, to generate the stereo content for being suitable to human eye viewing.
2nd, contrast verification is tested
Document [12] is not in the case where accounting for vision significance, directly to view picture stereo-picture quantitative study contrast
Influence of the factor to visual comfort, Fig. 8 are the comparison diagram of experimental result of the present invention and document [12] result, filled symbols
Line encloses zone of comfort and represents experimental result of the present invention, is denoted as RA, the lines of open symbols encloses zone of comfort and represents document
[12] experimental result, is denoted as RB, R6=RA∩RB, R7=RA\R6, R5=RB\R6。
Fig. 8 R6 regions show that zone of comfort obtained by document [12] is contained in zone of comfort obtained by the present invention, explanation substantially
The comfort level of marking area has stronger uniformity with the comfort level of view picture stereo-picture, is consistent with theory.R5 regions show,
When the contrast of left view point is very low, corresponding right viewpoint contrast is also very low in the zone of comfort that document [12] obtains,
But now the grain details of left and right view are all very fuzzy, brightness of image is low, and human eye is difficult to obtain by binocular fusion relatively comfortable
Stereo-picture, therefore R5 regions and do not meet human-eye visual characteristic obtained by document [12].And present invention gained zone of comfort
In, when left view point contrast is very low, right viewpoint contrast, can according to Fechner antinomys apparently higher than left view point contrast
Know, during binocular fusion, the larger viewpoint of contrast can make up the too small influence of another viewpoint contrast so that sense
The stereo-picture known is still in human eye vision comfort standard, therefore present invention gained zone of comfort more conforms to human eye vision
Characteristic.R7 regions show, for fixed left view, when comfort standard makes stereo content according to obtained by the present invention, right viewpoint
Contrast has bigger dynamic comfort standard, effectively relaxes the restrictive condition of stereo content making, is provided to make
It is convenient.
Because the more non-significant region of the marking area of stereo-picture includes more details and texture information, the sky of marking area
Between frequency concentrate on high frequency region[26].Contrast sensitivity function CSF is pointed out[27], dropped in the contrast sensitivity of high band human eye
It is low, identical contrast is perceived, it is necessary to increase the actual contrast value of image.Define CLThere is maximum relax for right viewpoint
Corresponding contrast, table 3 below list C in zone of comfort obtained by the present invention and document [12] to left view point just during suitable scopeL、
Desirable maximum, the minimum contrast C of left view pointLmax、CLmin, desirable maximum, the minimum contrast C of right viewpointRmax、CRminValue.
From table 3 it can be seen that present invention gained five characteristic values of zone of comfort are all than the height of document [12], and because quilt
The dominant eye of examination is right eye, CRmax、CRminIncreasing degree it is bigger, this phenomenon is consistent with human eye contrast degree sensitivity characteristic.Table 3 enters
One step illustrates that comfortable contrasted zones meet human-eye visual characteristic very much obtained by the present invention.
3rd, confirmatory experiment is tested
In order to verify that whether the present invention tests the obtained comfortable contrast matching figure of stereo-picture and disparity map with pervasive
Property, the present invention have chosen two width source stereo-pictures and carries out result verification in addition again, by this two width stereo-picture according in the present invention
State step and carry out contrast processing, each image obtains 441 width contrast distortion figures.Carry out subjective experiment and analyze experiment knot
Fruit.Toy this group of picture statistical averages pointThe qualified stereo-picture divided has 91 width, and left and right view contrast average value falls
Stereo-picture in the comfortable contrast matching figure of stereo-picture has 87 width, and proportion is 95.6% in qualified stereo-picture;
Equally it is calculated, accuracy is 97.5% in this group of picture of kid.The accuracy of two groups of pictures has been above 95%, demonstrates
Present invention experiment gained stereo-picture comfortably has preferable universality to degree matching figure and disparity map, therefore utilizes the present invention
Conclusion quickly directly can judge whether that image is comfortable according to the contrast of stereo-picture or so view, largely
On improve the present situation to waste time and energy of subjective experiment.
Bibliography
[1]Yano S,Emoto M,Mitsuhashi T.Two factors in visual fatigue caused
by stereoscopic HDTV images[J].Displays,2004,25(4):141-150
[2]Yano S,Ide S,Mitsuhashi T,et al.A study of visual fatigue and
visual comfort for 3D HDTV/HDTV images[J].Displays,2002,23(4):191-201.
[3]Lambooij M,Ijsselsteijn W,Heynderickx A I.Stereoscopic displays
and visual comfort:a review[M].2007
[4]Zheng H D,Yu Y J,Cheng W M.Areview on three-dimensional display
techniques. Optical Technique,2008,34(3):426
The bright three-dimensional stereo display techniques New research progress optical technologies of Zheng Huadong, Yu Yingjie, Cheng Wei, 2008,34 (3):
426
[5]Pan H,Daly S.3D video disparity scaling for preference and
prevention of discomfort[J]. Proceedings of SPIE-The International Society
for Optical Engineering,2011, 7863(4):786306-786306-8.
[6]Park J,Oh H,Lee S,et al.3D Visual Discomfort Predictor:Analysis of
Disparity and Neural Activity Statistics[J].IEEE Transactions on Image
Processing A Publication of the IEEE Signal Processing Society,2015,24(3):
1101.
[7]Yu Yang,Xutang Lib,Pin Xuc,et al.A Modifed Method for Multiple
Region Depth Mapping of Stereoscopic Image Generation[C].2016.IEEE
International Conference on Software Engineering and Service Science.
[8]LIU C,LI S M,ZHU D,et al.Quantitative research of the hue
parameter influence on the visual comfort of stereoscopic images.Journal of
Optoelectronics Laser,2014,01:178
Quantitative study photoelectricity of Liu Chang, Li Sumei, Zhu Dan, Zang Yan army the colourities to three-dimensional image vision comfort degree
Sub- laser, 2014,01:178
[9]ISO IWA 3 2005(E).Image safety:Reducing the incidence of
undesirable biomedical effects caused by visual image sequences.Switzerland:
ISO Copyright Office IHS, 2005
[10]Chen J,Zhou J,Sun J,et al.Visual discomfort prediction on
stereoscopic 3D images without explicit disparities[J].Image Communication,
2017,51(C):50-60.
[11]So G J,Kim S H,Kim J Y.Evaluation Model of the Visual Fatigue on
the 3D Stereoscopic Video[J].Acta Chirurgica Scandinavica Supplementum,2016,
325.
[12]LIU CH,LI S M.Measurement of contrast range affecting comfort of
stereoscopic images.Journal of Optoelectronics Laser,2014,25(4):748.
The measure photoelectron laser of the contrast range of Liu Chang, Li Su plum influence stereo-picture comfort levels, 2014,25
(4): 748.
[13]Jung C,Wang S.Visual comfort assessment in stereoscopic 3D images
using salient object disparity[J].Electronics Letters,2015,51(6):482-484.
[14]Harel,Jonathan,Christof K,et al.Graph-based visual
saliency.Advances in neural information processing systems,2007,19:545.
[15]Wang J,Da Silva M P,Le Callet P,et al.Computational model of
stereoscopic 3D visual saliency.IEEE Transactions on Image Processing,2013,22
(6):2151
[16]B,Platt J,Hofmann T.Graph-Based Visual Saliency[J]
.Advances in Neural Information Processing Systems,2007,19:545-552.
[17]Rhemann C,Hosni A,Bleyer M,et al.Fast cost-volume filtering for
visual correspondence and beyond[C].Computer Vision and Pattern Recognition
(CVPR),2011 IEEE Conference on.IEEE,2011:3017-3024.
[18]JIANG Q P,SHAO F,YU M,et al.An Objective Stereoscopic Image
Visual Comfort Assessment Metric Based on Visual Important Regions.Chinese
Journal of electronics and information technology,2014,36(04):875
Jiang Qiuping, Shao Feng, Jiang Gangyi, Yu Mei, Peng Zong lift the stereo image vision comfort level visitor of view-based access control model important areas
See evaluation method electronics and information journal, 2014,36 (04):875
[19]LI ZH M.Research on image segmentation algorithm based on Fuzzy
Clustering [Dissertation].Hunan University,2009.
Research on Algorithms for Image Segmentation [D] Hunan University of the Li Zhi plums based on fuzzy clustering, 2009.
[20]YU W J.Study of Otsu image segmentation based on Improved Genetic
Algorithm [Dissertation].Hainan University,2012.
Maximum between-cluster variance Research on Method of Image Segmentation [academic dissertation] the Hainan of the handsome of Yu Wen based on improved adaptive GA-IAGA
University, 2012.
[21]XIE Y B.Least square piecewise linear fitting.modern electric
power,1992,01:19.
Thank to the precious Sectional Line Fitting by means of Least Sequence Method modern electrics of friend, 1992,01:19.
[22]IVY Lab stereoscopic image database[OL].[2013-03-12].http://
ivylab.kaist.ac. kr/demo/3DVCA/3DVCA.html.
[23]Rec.ITU-R BT.1438.Subjective assessment of stereoscopic
television pictures [S].2000,1-14.
[24]Ma L S.Application of Grubbs test in equipment comparison test
[J].Opencast Mining Technology,2014,(04):62-64.
Application [J] the surface mining technologies of Ma Lishuan Grubbs' test methods in equipment Comparability test, 2014,
(04):62-64..
[25]Beijing Technology Quality Control Office.DBII/T384.5-2009,Image
information management system technical:image quality requirements and
evaluation methods.Beijing:China Standards Press,2009.
The part of Administration of Quality and Technology Supervision of Beijing .DBII/T384.5-2009 image information management systems technical specification the 5th:
Image quality requirements and evaluation method Beijing:China Standards Press, 2009
[26]Hou X,Zhang L.Saliency Detection:A Spectral Residual Approach
[C]//Computer Vision and Pattern Recognition,2007.CVPR'07.IEEE Conference
on.IEEE,2007:1-8.
[27]Johnson G M,Fairchild M D.A top down description of S‐CIELAB and
CIEDE2000[J].Color Research&Application,2010,28(6):425-435。
Claims (5)
1. a kind of stereo-picture comfort level contrast range measuring method based on salient region, it is characterized in that, step is as follows:
1) extraction of three-dimensional saliency map:Two dimensional image notable figure is combined to obtain three-dimensional saliency map with stereo image parallax figure;
2) the significantly acquisition of stereo-picture:Optimize processing to three-dimensional saliency map, three-dimensional saliency map through fuzzy theory and
Mask optimizes to obtain notable stereo-picture;
3) subjective experiment:Notable stereo-picture is subjected to contrast variation for step-length with 0.1 and carries out subjective experiment;
4) comfortable contrast matching figure and comfortable contrast disparity map are obtained.
2. the stereo-picture comfort level contrast range measuring method based on salient region as claimed in claim 1, it is special
Sign is that the extraction of three-dimensional saliency map comprises the concrete steps that:GBVS (Graph-based visual saliency) is used first
The plane notable figure SM of right view is calculated in algorithmR(x, y), then use Fast Stereo Matching Algorithm obtain using right view as
The disparity map d of benchmarkR(x, y), finally visual saliency map and right anaglyph are weighted to obtain solid using linear mode
Saliency map I (x, y), such as formula (1):
I (x, y)=w1dR(x,y)+w2SMR(x,y) (1)
In formula:I (x, y) is three-dimensional saliency map;w1And w2To weight proportion, and w1+w2=1.
3. the stereo-picture comfort level contrast range measuring method based on salient region as claimed in claim 1, it is special
Sign is that the acquisition detailed process of notable stereo-picture is that domain X is three-dimensional saliency map, by the pixel (x, y) in domain point
For two classes, marking area A and non-significant region B, A and B are an X divisions, i.e. A ∪ B=X,
The gray value of three-dimensional saliency map represents that the pixel in original image is under the jurisdiction of the degree of marking area, pixel (x,
Y) gray value is L, and this is to marking area A degree of membership
Need to carry out deblurring operation to obtain marking area A, mask images M (x, y) is obtained using the method for Threshold segmentation, such as
Shown in formula (2), segmentation threshold T is determined by maximum between-cluster variance method:
For any pixel point (x, y) in three-dimensional saliency map I (x, y), if (x, y) is more than to the degree of membership of marking area
Threshold value T, then the pixel belong to vision significance region, the white portion in corresponding mask images M (x, y), otherwise belong to black
Color region, mask image M (x, y) is optimized using morphologic opening and closing operation to obtain optimization mask image M'(x, y);
Finally optimization mask images are multiplied with original visual point image to obtain notable stereo-picture.
4. the stereo-picture comfort level contrast range measuring method based on salient region as claimed in claim 1, it is special
Sign is that contrast variation carries out comprising the concrete steps that for subjective experiment, after obtaining notable stereo-picture, using contrast linear transformation
Method, different grades of contrast distortion is carried out to left and right view, obtains a series of stereo-picture to be measured of distortions, image is carried out
During contrast variation, the gray value of note image original image vegetarian refreshments is In, and conversion after image is calculated by contrast variation's formula (3)
The gray value Out of vegetarian refreshments:
Out=Avg+ (In-Avg) × (1- α) (3)
Wherein Avg represents the average gray value of whole pictures, and α ∈ [- 1,1] are the setting contrast factor, controls output pixel point
Relation with inputting pixel gray value, the conversion step-length for taking α is 0.1, obtains 21 contrast variation's straight lines, corresponding 21 kinds etc.
The contrast distortion of level, transverse axis are the gray value In of input pixel, and the longitudinal axis is that pixel spends after conversion obtained ash by contrast
Angle value Out, every contrast variation's straight line all passing points (Avg, Avg) and on (Avg, Avg) symmetrically, ensure that contrast become
The mean flow rate for changing front and rear picture keeps constant.
5. the stereo-picture comfort level contrast range measuring method based on salient region as claimed in claim 1, it is special
Sign is to obtain comfortable contrast matching figure and comfortable contrast disparity map comprises the concrete steps that, is intended using least square piecewise linearity
It is legal that 4 groups of experimental datas are handled, obtain 4 groups of stereo-pictures and comfortably normalize contrast matching figure MM (Matching
Map contrast disparity map DM (Difference Map)) and is comfortably normalized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710729853.2A CN107680076A (en) | 2017-08-23 | 2017-08-23 | Stereo-picture comfort level contrast range measuring method based on salient region |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710729853.2A CN107680076A (en) | 2017-08-23 | 2017-08-23 | Stereo-picture comfort level contrast range measuring method based on salient region |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107680076A true CN107680076A (en) | 2018-02-09 |
Family
ID=61134411
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710729853.2A Pending CN107680076A (en) | 2017-08-23 | 2017-08-23 | Stereo-picture comfort level contrast range measuring method based on salient region |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107680076A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108521568A (en) * | 2018-03-26 | 2018-09-11 | 天津大学 | The assay method of stereo-picture comfort level disparity range based on salient region |
CN109727274A (en) * | 2018-11-28 | 2019-05-07 | 吉林大学 | Method for processing video frequency |
CN116664431A (en) * | 2023-05-30 | 2023-08-29 | 原朗智芯(苏州)数字科技有限公司 | Image processing system and method based on artificial intelligence |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120121138A1 (en) * | 2010-11-17 | 2012-05-17 | Fedorovskaya Elena A | Method of identifying motion sickness |
CN103096125A (en) * | 2013-02-22 | 2013-05-08 | 吉林大学 | Stereoscopic video visual comfort evaluation method based on region segmentation |
CN103139598A (en) * | 2013-03-14 | 2013-06-05 | 天津大学 | Measuring method of normalization contrast ranges affecting comfort levels of three-dimensional pictures |
CN104853183A (en) * | 2015-05-22 | 2015-08-19 | 天津大学 | Objective evaluation method for perceiving three-dimensional sense of three-dimensional image |
CN105744256A (en) * | 2016-03-31 | 2016-07-06 | 天津大学 | Three-dimensional image quality objective evaluation method based on graph-based visual saliency |
CN106993183A (en) * | 2017-03-28 | 2017-07-28 | 天津大学 | Comfortable brightness quantitative approach based on stereo-picture marking area |
-
2017
- 2017-08-23 CN CN201710729853.2A patent/CN107680076A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120121138A1 (en) * | 2010-11-17 | 2012-05-17 | Fedorovskaya Elena A | Method of identifying motion sickness |
CN103096125A (en) * | 2013-02-22 | 2013-05-08 | 吉林大学 | Stereoscopic video visual comfort evaluation method based on region segmentation |
CN103139598A (en) * | 2013-03-14 | 2013-06-05 | 天津大学 | Measuring method of normalization contrast ranges affecting comfort levels of three-dimensional pictures |
CN104853183A (en) * | 2015-05-22 | 2015-08-19 | 天津大学 | Objective evaluation method for perceiving three-dimensional sense of three-dimensional image |
CN105744256A (en) * | 2016-03-31 | 2016-07-06 | 天津大学 | Three-dimensional image quality objective evaluation method based on graph-based visual saliency |
CN106993183A (en) * | 2017-03-28 | 2017-07-28 | 天津大学 | Comfortable brightness quantitative approach based on stereo-picture marking area |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108521568A (en) * | 2018-03-26 | 2018-09-11 | 天津大学 | The assay method of stereo-picture comfort level disparity range based on salient region |
CN109727274A (en) * | 2018-11-28 | 2019-05-07 | 吉林大学 | Method for processing video frequency |
CN116664431A (en) * | 2023-05-30 | 2023-08-29 | 原朗智芯(苏州)数字科技有限公司 | Image processing system and method based on artificial intelligence |
CN116664431B (en) * | 2023-05-30 | 2024-04-12 | 新疆美特智能安全工程股份有限公司 | Image processing system and method based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110991266B (en) | Binocular face living body detection method and device | |
Park et al. | 3D visual discomfort prediction: vergence, foveation, and the physiological optics of accommodation | |
CN103986925B (en) | based on the stereoscopic video visual comfort evaluation method of luminance compensation | |
CN104853184B (en) | Detection method for parallax scope influencing comfort level of stereo image | |
CN102497565B (en) | Method for measuring brightness range influencing comfort degree of stereoscopic image | |
CN108449596A (en) | A kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level | |
CN107680076A (en) | Stereo-picture comfort level contrast range measuring method based on salient region | |
CN107194927A (en) | The measuring method of stereo-picture comfort level chromaticity range based on salient region | |
CN104853185A (en) | Stereo video comfort evaluation method combining multiple parallaxes with motion | |
CN102595185A (en) | Stereo image quality objective evaluation method | |
CN106973288A (en) | A kind of three-dimensional video-frequency Comfort Evaluation method and device | |
CN108521568A (en) | The assay method of stereo-picture comfort level disparity range based on salient region | |
CN105976351A (en) | Central offset based three-dimensional image quality evaluation method | |
CN107452032A (en) | Human body back depth image preprocess method | |
CN106993183A (en) | Comfortable brightness quantitative approach based on stereo-picture marking area | |
CN103139598B (en) | Measuring method of normalization contrast ranges affecting comfort levels of three-dimensional pictures | |
Chen et al. | Study of subject agreement on stereoscopic video quality | |
CN106934770A (en) | A kind of method and apparatus for evaluating haze image defog effect | |
Wan et al. | Depth perception assessment of 3D videos based on stereoscopic and spatial orientation structural features | |
Kim et al. | Quality assessment of perceptual crosstalk on two-view auto-stereoscopic displays | |
Barkowsky et al. | Is visual fatigue changing the perceived depth accuracy on an autostereoscopic display? | |
CN107483918A (en) | Stereo image quality evaluation method is referred to based on conspicuousness entirely | |
CN108259888A (en) | The test method and system of stereo display effect | |
Wu et al. | Subjective quality assessment of stereo image compression based on stereoscopic fusion in binocular vision | |
CN110060291B (en) | Three-dimensional apparent distance resolving method considering human factors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180209 |